The invisible AI

The AI that works in luxury is the one you don't see

$42 million in EBITDA. The RealReal's first profitable fiscal year ever. And the AI behind this result never had a single conversation with a customer.

My conviction: the next wave of value creation in luxury won't come from customer-facing conversational agents, but from invisible infrastructure -- pricing, authentication, forecasting, operations. The numbers published in recent weeks confirm this spectacularly.

The RealReal -- when operational AI delivers results

The RealReal, the leading American platform for authenticated luxury resale -- which I already discussed last week as an involuntary training corpus for LLMs (Large Language Models) -- has just published its annual results. GMV (Gross Merchandise Value) of $2.13 billion in 2025 (+16%), EBITDA of $42 million -- the first positive full fiscal year in the platform's history. Revenue grew 11% to $600 million.

The engine behind this profitability: Athena. This AI system processes 35% of incoming units -- sorting, categorization, pricing, listing. It compresses the intake-to-listing cycle (the time between receiving an item and putting it up for sale), the model's historical bottleneck. In parallel, SmartSales, a predictive tool, identifies potential consignors in the field -- clients likely to entrust pieces for resale. None of this is visible to the end customer. Zero chatbot. Zero interface. Everything is in the machine.

And the houses whose pieces feed the platform -- Chanel, Hermes, Louis Vuitton are among the most consigned brands on The RealReal -- are indirectly funding this model. Each bag authenticated by Athena trains the algorithm on the manufacturing codes of a house that never consented to sharing this data. The paradox is complete. And nobody is talking about it.

This is the exact opposite of the dominant narrative. We're being sold customer-facing conversational AI. The one transforming margins is the one talking to the warehouse. Daniel Langer, luxury analyst and Luxury Daily columnist, put it this way this week: "Luxury brands that respond to the AI revolution by automating their customer experience are investing to eliminate precisely what justifies their premium." The RealReal didn't automate its customer experience. It automated its warehouse.

Luxury invests in "run," not in "change"

Luca de Meo, Kering's new CEO whose plan we detail in WHAT'S MOVING below, delivers the diagnosis bluntly: "In luxury, we probably invest around 1% of revenue." He comes from automotive -- 10%. The Bain x Comite Colbert report (4th edition, Sept. 2025) refines the picture: houses allocate an average of 3.1% of revenue to technology -- and 63% of that budget goes to "run," maintaining existing systems. Only 37% goes to "change," meaning transformation. The Deloitte Global Powers of Luxury 2026 report (420 C-suite executives, 10 countries, detailed in WHAT'S MOVING) confirms the disconnect: customer experience and loyalty are identified as the number one opportunity (28.6%), but actual investment continues to flow toward maintaining the status quo. Putting 63% into maintenance when The RealReal demonstrates that operational AI is what produces EBITDA is a risky bet.

I believe that houses continuing to over-fund "run" are making a strategic error: they're maintaining systems that produce no competitive advantage, while players like The RealReal or Profound are building the infrastructure of the future.

Context: agentic commerce is being built without luxury

James Cadwallader, co-founder of Profound whose fundraise we analyzed in THE SIGNAL, sums it up in one sentence: "The front door of the internet is changing for the first time in 25 years." The State of Fashion 2026 (BoF -- Business of Fashion -- and McKinsey), which we source in WHAT'S MOVING below, puts numbers on the movement: the potential of agentic commerce is estimated at $3 to $5 trillion by 2030. Shopping searches on AI platforms surged +4,700% in one year. The customer journey is shifting from "human-first" to "agent-first" across six stages -- from the initial prompt to post-sale support. But the houses that will survive this wave won't be those with the best conversational assistant. They'll be those whose foundation -- structured product data, dynamic pricing, authentication, forecasting -- is solid enough to feed any agent, internal or external. Agentic commerce rewards brands that have rich descriptions, clear policies, clean data. Machine criteria, not desire criteria -- and yet, in a world of shopping agents, these are the ones that decide.

And Deloitte's data on customer trust -- which we detail in WHAT'S MOVING below -- confirms what the field teaches: when the customer knows AI is involved, trust drops. In luxury, discretion is not a technical constraint. It's an invariant.

The RealReal accomplished in two years what most houses haven't even started: turning operational AI into financial results. The question is no longer whether AI works in luxury. It's why it only works where nobody can see it.